The work that is hardest to automate is the work we have historically valued least. That paradox is about to become very expensive.
The care economy — the sector encompassing childcare, eldercare, healthcare support, social work, and disability services — employs tens of millions of people in every wealthy country. It is the infrastructure on which most other economic activity rests. Without functional care systems, parents can't work, the elderly can't age with dignity, and the sick can't recover. And yet the median hourly wage for a childcare worker in the United States is $14.60. For a home health aide, it's $15.25.
Defining the Care Economy
"Care economy" is a broad term that covers both paid and unpaid care work. The unpaid portion — estimated by the International Labour Organization to represent approximately $11 trillion in annual economic value globally if compensated at minimum wage — consists overwhelmingly of work done by women: raising children, caring for elderly parents, supporting ill family members.
The paid care sector includes:
- Early childhood education and childcare: Nursery teachers, daycare workers, childminders
- Healthcare support: Home health aides, personal care workers, nursing assistants
- Social services: Social workers, case managers, family support workers
- Disability services: Support workers, residential care staff
What unites these roles is the centrality of human relationship to the work itself. Care work cannot, in its essence, be performed at arm's length. It requires physical presence, attentiveness to individual need, emotional attunement, and the sustained establishment of trust. These are not merely soft attributes. They are the functional requirements of the work.
Why It Pays So Little
The low wages in care work are not an accident. They are the outcome of intersecting structural factors that have been well-documented by feminist economists for decades.
The devaluation of feminized labor. Care work is predominantly performed by women, and work predominantly performed by women has historically been compensated below its economic value. This is not a subtle pattern. Studies that have controlled for education, experience, and job demands consistently find a wage penalty associated with working in female-dominated occupations — independent of the characteristics of the workers themselves.
The "public good" problem. Many care services produce benefits that extend far beyond the individual receiving care. Quality early childhood education produces long-term returns in child development, educational attainment, and reduced social costs — but most of those returns accrue to society broadly, not to the employer of the childcare worker. This creates a systematic tendency toward underinvestment in care quality and undercompensation of care workers.
Market structure. Care services are frequently provided in highly fragmented, low-margin markets where providers compete primarily on price. Families paying for childcare are often at or near the limit of what they can afford; facilities operate on thin margins; workers bear the cost in wages. This is a market that requires structural subsidy to function well — subsidy that most countries provide inadequately.
What AI Changes (and Doesn't)
The AI automation wave is doing something counterintuitive to the care economy: it's increasing its relative importance while doing almost nothing to solve its underlying problems.
The jobs that AI is most effectively displacing are knowledge-intensive, routine-cognitive tasks — data processing, document analysis, customer service, basic legal and financial work. These are precisely the jobs that do not require sustained human presence and emotional attunement. Care work, by contrast, is highly resistant to automation precisely because its value is intrinsically relational.
A 2023 analysis by the McKinsey Global Institute rated personal care and service occupations among the lowest in automation potential. An elderly person with dementia requires a care worker who can read their emotional state moment to moment, adapt to unexpected behavior, and provide reassurance that is responsive to the specific person. No current AI system does this.
The result is a widening gap: as automation displaces higher-paid knowledge work, the relative share of the economy represented by hard-to-automate care work grows — but the compensation structure remains unchanged. We are moving toward an economy where care work is proportionally more important and still structurally undercompensated.
The Demographic Accelerant
Overlaid on the AI transition is a demographic reality that is already reshaping labor markets across wealthy countries.
Japan, Germany, South Korea, Italy, and the United States are all facing sharp increases in the over-65 population over the next two decades. The demand for eldercare specifically — home health aides, nursing assistants, dementia care specialists — is projected to grow faster than almost any other occupational category in this period.
The United States Bureau of Labor Statistics projects that home health and personal care aide roles will add more jobs between 2022 and 2032 than any other single occupation. The median wage for these roles remains under $16 per hour.
A society is building an enormous demand for work it has not figured out how to adequately compensate. The care shortage that would result — not from automation displacing workers, but from inadequate compensation preventing worker entry and retention — is already visible in eldercare settings in multiple countries.
The Policy Problem
The care economy presents a genuine policy challenge that does not have a clean market solution.
Raising care wages to market-clearing levels — levels that attract and retain sufficient workers — would require either dramatically increasing what families pay for care (pricing many out of the market) or dramatically increasing public subsidy for care services. Both options face significant political obstacles.
What is not a viable option is continuing the current system unchanged while demographic and technological pressure accumulates. The care economy will not fix itself. The question is whether policy intervention comes before or after the shortage becomes acute enough to force it.
The paradox of the care economy — most important, least rewarded — is not just an equity problem. It's a structural vulnerability. And the AI transition is making it more visible precisely by revealing what machines can't replace.